First Week Review of CarND Program
--
So its been one week since the program started. I tried to commit by spending 1–2 hours per day before going to bed to learn the lesson. It is only the beginning but already learn quite much related to computer vision. The given method and material are very straight forward, easy to follow. For somebody like me who never learn computer vision before, IMHO, the way they presented material are very well. I don’t have to understand the complex theory behind several technique such as canny edge detection, or hough transformation. They give the code template (in python) to try, let me play with some parameters to see how’s the output. By using that method I can get a grasp of how the algorithm work without understanding the theory behind it. For example, below grayscaled image.
Converted into image with canny edge detection algorithm.
The script template already provided, all I need to do just tweak parameters to get expected result.
Once comfortable performing several trial-error to tweak parameters, a link given in order to learn deeper of mentioned technique or algorithm. For example, topic related to Hough-Transformation can be read from this blog.
The first one week material are related to computer vision that contain topic such as color selection, region masking, canny-edge detection, and hough transformation.
Project
In term-1 (first 3 months) there will be several project to submit. First project deadline is one-week only. The first project goal is to detect lane lines in a road technique related to computer vision. Each project must be submitted according to its deadline. If we can’t submit as per deadline and reach end of first term, then we must join next cohort (repeat the class) and lost opportunity to be included on recruitment process (if interested to find job within autonomous car industry).
The project basically designed to use knowledges given on the lesson itself. We must be able to let the computer detect lane lines. We can try on one image first, and then on the video.
Above image contain the lane lines, and our task is to create pipeline code to detect the lane lines.
Once it is success for one image, we can deploy by using a video file as its input as target for this project. Video itself basically a series of images. This is the sample original video.
Here’s the result
The code must be able to detect lane lines of different color as well (that’s why canny-edge used on this case).
The code also must be able to detect whenever lane lines not straight i.e curve. That is one of reason hough-transformation used on this case: to transform the lines that is originally composed by (x,y) coordinate into hough-space so when the shape of lane changed the computer still able to detect it.
The project submission not only containing code, but also the write-up of our approach how to produce expected result. Udacity team want to make sure that we’re comfortable with common tools used by developers, so the expectation is to use markdown format — although we can use .pdf as well.
We can submit the project by giving the repo link in github to be reviewed, or upload the zipped files.
It took few hours for the reviewers from Udacity to review my submission. Once finished, they will send email confirmation and we can see the result. Here’s my project review result for first project.
Career Orientation and Opportunities
Nanodegree program designed to prepare the trainee to have adequate knowledge or skills to be applied immediately on the industry, especially their partner company. Beside giving technical material, udacity also giving lesson related to career orientation, as well as lesson related to job search strategies or composing resume. So this program is very good for fresh-graduate students, or people who want to switch career into different field.
Udacity’s Program
One week has passed, remains 11 weeks to learn various interesting stuff. I take a peek into several lesson ahead and all of them seems juicy :)
One of my colleague once asked if the program also suitable for him. IMHO, every person have different goal on learning. If you’re not interested on artificial intelligence and autonomous cars then why wasting money on it.
As per nanodegree program itself, Udacity have quite long list of course that you can check here.
If, for example, you’re interested to learn VR (Virtual Reality) because curious to know how the technology work then go with it. Review of each course also available so we can decide to purchase it or not.
As per Self-Driving Cars NanoDegree program itself, so far I am very satisfied, so I would recommend anyone who interested in such technology to join the program. Basic programming skills required? IMHO, yes. Although for some people maybe not required, but at least you have to prepare the environment to test the code. I don’t have issue so far because python is part of language I used daily, but in case you have difficulties, the Slack chat room and also Forum will be available. People are exchanging information and help each others to solve other students who might have difficulties i.e preparing they correct environment. Moreover, every student got one dedicated mentor in case they find problem during the course. So nothing to worry.
Cheers.
